1:00 PM - 1:15 PM
▲ [20p-D104-1] Quantification of spin drift in heavily-doped Si nonlocal devices
Keywords:silicon spintronics, nonlocal spin-transport device, spin drift
The maturity of Si-based technology provides compelling motivation to integrate the spin functionality in novel Si-based devices. Recently, we demonstrated the creation of a giant spin accumulation in Si up to room temperature using the nonlocal (NL) scheme [1]. In the latter study, spin transport was studied in the low bias regime, where it can be described as purely diffusive using the standard theory for spin injection and diffusion. In the high bias regime, however, the presence of an electric field (E) in the Si channel causes spin drift that can strongly affect spin transport and spin accumulation voltages in Si [2,3]. Here, we study the spin transport in heavily-doped n-type Si NL devices in the presence of E in the high bias regime and quantify the effect of spin drift on the NL spin signals.
The epitaxial Fe/MgO magnetic tunnel contacts were deposited by molecular beam epitaxy on a SOI substrate having a 70 nm-thick n-type Si(001) channel with a phosphorous concentration of 2.4 × 1019 cm-3. The NL structures consist of two central ferromagnetic electrodes (FM1 and FM2) separated by a spacing d and two outer non-magnetic Au/Ti reference contacts. To study the spin-drift effect in the NL geometry, we use two different measurement configurations called NL1 and NL2, in which the direction of the electric field E is reversed, while the detector remains unbiased. To vary the strength of the electric field, we designed devices with different contact widths of FM1 and FM2 on the same sample.
We first measure the NL spin-valve and Hanle signals in the low bias regime and extract a spin lifetime of 13 ns, a spin-diffusion length of 1.7 mm, and a tunnel spin polarization of Fe/MgO of 40 % at 60 K, in agreement with our previous work [1]. We then perform NL spin-valve measurements using both NL1 and NL2 schemes and different injector sizes at various biases. We show that for spin injection i) the ratio of the spin signal NL2/NL1 is enhanced by E and ii) the spin-transport length with E is larger or smaller than the spin-diffusion length, depending on whether the transport is up-stream or down-stream. Notably, we find that although the theory of Ref. 2 predicts the correct trends, it overestimates the effect of E.
The epitaxial Fe/MgO magnetic tunnel contacts were deposited by molecular beam epitaxy on a SOI substrate having a 70 nm-thick n-type Si(001) channel with a phosphorous concentration of 2.4 × 1019 cm-3. The NL structures consist of two central ferromagnetic electrodes (FM1 and FM2) separated by a spacing d and two outer non-magnetic Au/Ti reference contacts. To study the spin-drift effect in the NL geometry, we use two different measurement configurations called NL1 and NL2, in which the direction of the electric field E is reversed, while the detector remains unbiased. To vary the strength of the electric field, we designed devices with different contact widths of FM1 and FM2 on the same sample.
We first measure the NL spin-valve and Hanle signals in the low bias regime and extract a spin lifetime of 13 ns, a spin-diffusion length of 1.7 mm, and a tunnel spin polarization of Fe/MgO of 40 % at 60 K, in agreement with our previous work [1]. We then perform NL spin-valve measurements using both NL1 and NL2 schemes and different injector sizes at various biases. We show that for spin injection i) the ratio of the spin signal NL2/NL1 is enhanced by E and ii) the spin-transport length with E is larger or smaller than the spin-diffusion length, depending on whether the transport is up-stream or down-stream. Notably, we find that although the theory of Ref. 2 predicts the correct trends, it overestimates the effect of E.